Bridgewater: Open-weight model outperforms GPT and Claude in finance tests

The fact

Internal tests indicate that the correct answers were not publicly available, reducing the effectiveness of commercial models.

The study highlights the limitations of general-purpose LLMs in specialized tasks like finance, despite their capabilities.